Annular visual inspection robot for detecting inner surface defects of oil and gas pipeline

Abstract

The invention relates to an annular visual inspection robot for detecting defects on the inner surface of an oil and gas pipeline, which belongs to the technical field of pipeline inspection and comprises an imaging unit, a processing unit, a battery unit and an odometer wheel unit, the imaging unit, the processing unit, the battery unit and the odometer wheel unit are connected in series; the imaging unit is used for shooting oil-gas pipeline inner surface images; the processing unit is used for performing defect detection according to the shot inner surface image of the oil and gas pipeline; the battery unit is used for supplying power to the imaging unit and the processing unit; the odometer wheel unit is used for recording movement mileage; the imaging unit comprises a camera, a conical mirror and a glass cylinder; the glass cylinder seals the imaging unit, a containing space is formed in the glass cylinder, the conical mirror is arranged in the containing space, and external light penetrates through the glass cylinder, reaches the conical mirror and is projected into the camera through the conical mirror, so that an annular image of the inner surface of the oil and gas pipeline is shot through the camera. The visual field range and the image quality of the robot in the complex pipeline environment are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of pipeline detection, and particularly relates to an annular vision detection robot for detecting internal surface defects of oil and gas pipelines. Background Art

[0002] The internal detection of oil and gas pipelines is crucial for ensuring pipeline safety. As a key infrastructure for oil and gas transportation, the safe operation of pipelines is directly related to the stability of energy supply, environmental protection, and the safety of public life and property. However, due to the long-term exposure of oil and gas pipelines to complex working environments, they are internally threatened by various factors such as corrosion, cracks, welding defects, and foreign object accumulation. If these hidden dangers are not discovered and handled in a timely manner, it is extremely easy to cause major accidents such as pipeline leaks and explosions. Therefore, carrying out efficient and accurate internal detection is an important technical guarantee for ensuring the safe operation of pipelines, and at the same time, it is also one of the core means to avoid environmental pollution and economic losses.

[0003] However, traditional detection methods are inefficient and inaccurate, especially in high-pressure and complex environments. Existing technologies for collecting internal pipeline images of small-diameter pipelines, such as CCTV (Closed Circuit Television), etc., mostly use depth-of-field imaging methods. In terms of imaging, depth-of-field imaging has problems such as limited field of view, image clarity and resolution issues, and image distortion. In addition, internal pipeline detection is often affected by factors such as changes in pipe diameter, complex media, and insufficient light, resulting in unstable image quality. Therefore, it is difficult for existing technologies to ensure equipment stability and image quality in high-pressure and complex environments. Summary of the Invention

[0004] To solve the above problems of the existing technology, the present invention provides an annular vision detection robot for detecting internal surface defects of oil and gas pipelines.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention provides an annular vision detection robot for detecting internal surface defects of oil and gas pipelines, including: an imaging unit, a processing unit, a battery unit, and a odometer wheel unit;

[0007] The imaging unit, the processing unit, the battery unit, and the odometer wheel unit are connected in series;

[0008] The imaging unit is used for taking images of the internal surface of the oil and gas pipeline;

[0009] The processing unit is used for detecting defects based on the taken images of the internal surface of the oil and gas pipeline;

[0010] The battery unit is used for supplying power to the imaging unit and the processing unit;

[0011] The mileage wheel unit is used to record the movement mileage;

[0012] The imaging unit includes a camera, a conical mirror, and a glass cylinder;

[0013] The glass cylinder seals the imaging unit. An accommodation space is formed inside the glass cylinder. The conical mirror is arranged in the accommodation space. External light passes through the glass cylinder and reaches the conical mirror, and is projected onto the camera through the conical mirror, so as to capture a circular image of the inner surface of the oil and gas pipeline through the camera.

[0014] The beneficial effects of the present invention are reflected in:

[0015] The present invention provides a circular vision detection robot for detecting inner surface defects of oil and gas pipelines. It adopts the conical mirror reflection ring imaging technology, captures high-resolution images reflected by the conical mirror surface through a camera, realizes a 360° panoramic clear image of the inner wall of the pipeline, reduces image distortion, ensures the accurate restoration of the defect area, improves the vision range and image quality of the robot in a complex pipeline environment, and further improves the accuracy of defect detection based on visual images. Description of the Drawings

[0016] Figure 1 It is a schematic three-dimensional structure diagram of a circular vision detection robot for detecting inner surface defects of oil and gas pipelines provided by the present invention.

[0017] Figure 2 It is a schematic three-dimensional structure diagram of another circular vision detection robot for detecting inner surface defects of oil and gas pipelines provided by the present invention.

[0018] Figure 3 It is a schematic structure diagram of a differential pressure rotary cleaning unit provided by the present invention.

[0019] Figure 4 It is a schematic flow diagram of a method for detecting inner surface defects of oil and gas pipelines provided by the present invention.

[0020] Description of the Reference Numerals:

[0021] 1 - Imaging unit, 11 - Camera, 12 - Conical mirror, 13 - Glass cylinder, 2 - Processing unit, 3 - Battery unit, 4 - Mileage wheel unit, 41 - Mileage wheel, 42 - Mileage wheel base, 43 - Mileage wheel bracket, 5 - Sealing head, 6 - Leather cup drive unit, 7 - Differential pressure rotary cleaning unit, 71 - Base, 72 - Rotary fan blade, 73 - Rotary glass cylinder, 711 - Air inlet hole, 712 - Air outlet hole. Detailed Embodiments

[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] The specific embodiments provided by the present invention are as follows:

[0024] Referring to the attached Figure 1 figures, a three-dimensional structural schematic diagram of an annular vision inspection robot for detecting inner surface defects of oil and gas pipelines provided by an embodiment of the present invention is shown.

[0025] Referring to the attached Figure 2 figures, another three-dimensional structural schematic diagram of an annular vision inspection robot for detecting inner surface defects of oil and gas pipelines provided by an embodiment of the present invention is shown.

[0026] An embodiment of the present invention provides an annular vision inspection robot for detecting inner surface defects of oil and gas pipelines, including: an imaging unit 1, a processing unit 2, a battery unit 3, and a mileage wheel unit 4.

[0027] The imaging unit 1, the processing unit 2, the battery unit 3, and the mileage wheel unit 4 are connected in series.

[0028] The imaging unit 1 is used to capture images of the inner surface of the oil and gas pipeline.

[0029] The processing unit 2 is used to detect defects based on the captured images of the inner surface of the oil and gas pipeline.

[0030] The battery unit 3 is used to supply power to the imaging unit 1 and the processing unit 2.

[0031] The mileage wheel unit 4 is used to record the movement mileage.

[0032] The imaging unit 1 includes a camera 11, a conical mirror 12, and a glass cylinder 13.

[0033] The glass cylinder 13 seals the imaging unit 1. A receiving space is formed inside the glass cylinder 13. The conical mirror 12 is arranged in the receiving space. External light passes through the glass cylinder 13 and reaches the conical mirror 12, and is projected onto the camera 11 through the conical mirror 12, so as to capture annular images of the inner surface of the oil and gas pipeline through the camera 11.

[0034] Optionally, the inclination angle of the conical mirror 12 is 45°.

[0035] Through the above embodiments, the annular vision inspection robot of the present invention can effectively adapt to inspection tasks under high pressure and complex environments. Through reasonable equipment layout and pressure-resistant protection design, high-quality image acquisition and stable equipment operation are achieved, improving the efficiency and accuracy of the detection of internal surface defects in oil and gas pipelines. The design of the odometer wheel further enhances the navigation and positioning capabilities of the robot, making the detection process more precise and reliable.

[0036] The beneficial effects of the present invention are reflected in:

[0037] The present invention provides an annular vision inspection robot for detecting internal surface defects in oil and gas pipelines. It adopts the imaging technology of conical mirror reflection ring. By capturing high-resolution images reflected by the conical mirror surface through a camera, a 360° panoramic clear image of the pipeline inner wall is realized, reducing image distortion, ensuring the accurate restoration of the defect area, enhancing the vision range and image quality of the robot in a complex pipeline environment, and further improving the accuracy of defect detection based on visual images.

[0038] In a possible embodiment, both the conical mirror 12 and the camera 11 are centered.

[0039] In the present invention, by centering both the conical mirror 12 and the camera 11, the symmetry of the optical system can be ensured, maximizing the 360° panoramic coverage, effectively reducing local distortion and distortion of the image, and at the same time optimizing the optical path transmission efficiency, thereby improving the imaging quality and the accuracy of defect detection.

[0040] In a possible embodiment, the glass cylinder 13 is made of sapphire material.

[0041] It should be noted that the glass cylinder 13 made of sapphire material can ensure that the system can withstand a pressure of 10 MPa, and light can pass through the cylinder wall to reach the conical surface for reflection.

[0042] In the present invention, using sapphire material as the glass cylinder 13 can provide excellent pressure resistance, can withstand a pressure of up to 10 MPa, ensuring the safe operation of the equipment in a high-pressure environment. At the same time, sapphire material has extremely high optical transmittance, can effectively transmit light to the conical mirror, ensure the imaging quality, and has excellent corrosion resistance, adapting to complex pipeline detection environments.

[0043] In a possible embodiment, the processing unit 2 includes: a computer, a power control board, and a delay switch board. The power control board is connected to the battery unit 3. The power control board and the imaging unit 1 are both connected to the computer. The imaging unit 1, the odometer wheel unit 4, the LED control board, and the LED lights are all connected to the power control board. The LED control board is used to trigger the supplementary lighting function of the LED supplementary light, which can optimize energy consumption.

[0044] Furthermore, the processing unit 2 may further include a supercapacitor, which is used to provide instantaneous high-current output to ensure the power supply stability of the device during high-load operation.

[0045] Further, the battery is first connected to the switch circuit, and the load end of the switch circuit is connected to the power supply part of the power supply circuit. The power supply circuit is responsible for providing the required voltages for the odometer wheel, the camera optocoupler input, the industrial camera, and the pocket computer. Specifically, the odometer wheel and the camera optocoupler input obtain a 5V voltage, and the industrial camera and the pocket computer obtain 12V voltages respectively. To prevent the pocket computer from shutting down due to a decrease in the battery voltage during the lighting of the LED lamp, a supercapacitor is connected in parallel across the battery to provide instantaneous high-current output and ensure the power supply stability of the device during high-load operation.

[0046] In the present invention, by configuring a delay switch board in the processing unit 2, the time-sharing startup of each component of the device can be realized, avoiding the impact of high load at startup on the battery, thereby optimizing energy consumption, prolonging the battery life, and improving the stability of system operation. Especially in a complex pipeline detection environment where multiple modules operate in parallel, it ensures the reliable and efficient operation of the device.

[0047] In a possible implementation manner, the odometer wheel unit 4 includes: an odometer wheel 41, an odometer wheel base 42, an odometer wheel bracket 43, a rotating shaft, a magnet, and a magnetic encoder. The odometer wheel bracket 43 is connected to the odometer wheel base 42, and the odometer wheel 41 is arranged on the odometer wheel bracket 43 through the rotating shaft. A radially magnetized magnet is installed on the rotating shaft, and the magnet is located at the geometric center of the rotating shaft. When the annular vision detection robot runs inside the pipeline, the odometer wheel 41 contacts the inner wall of the pipeline and drives the rotating shaft connected by a key to rotate. When the rotating shaft rotates, the magnet rotates with the rotating shaft, and the magnetic encoder senses the position change of the magnet and generates a pulse signal. The camera 11 unit is triggered to take pictures through the pulse signal to ensure that the images at different positions inside the pipeline can be accurately recorded and analyzed.

[0048] Specifically, the pulse signal is then sent to the optocoupler input of the industrial camera, and the optocoupler input outputs a photographing pulse signal according to the received number of pulses and a preset number of pulses.

[0049] In the present invention, by adopting the design of a magnet and a magnetic encoder in the odometer wheel unit 4, the rotation position of the odometer wheel can be sensed in real time, and the camera 11 is accurately triggered to take pictures through the generated pulse signal. This can ensure that the images taken at different positions inside the pipeline correspond to their specific positions one by one, improving the positioning accuracy and the spatial consistency of the image data during the detection process, thereby realizing more accurate defect recording and analysis.

[0050] In a possible implementation, the imaging unit 1 further includes an LED fill light, and the fill light function of the LED fill light is triggered by the mileage wheel unit 4 and the LED control board.

[0051] Optionally, the LED light board is installed on the same plane as the industrial camera lens, and the fill light function is realized through the mileage wheel trigger, LED control board and power supply circuit to ensure clear images in low-light environments.

[0052] Specifically, when the optocoupler input collects a preset number of pulse signals, that is, when the mileage wheel runs a fixed mileage, the industrial camera will send a photo pulse signal to the power circuit. After the LED control board receives this pulse signal, the control battery is directly connected to the LED light board, so that the light board lights up. At the same time, the industrial camera starts to collect images of the conical mirror to ensure that clear images of the inner wall of the pipeline can be captured even in low-light environments.

[0053] In the present invention, by integrating an LED fill light in the imaging unit 1 and triggering its fill light function by the mileage wheel unit 4, accurate and synchronous lighting can be provided in low-light or no-light environments to ensure imaging quality. At the same time, the fill light is intelligently controlled by using an optical coupler and an LED control board so that it is only lit when shooting is required, effectively reducing energy consumption, extending battery life, and ensuring the clarity of the image of the inner wall of the pipeline and the reliability of detection.

[0054] In a possible implementation, the annular visual inspection robot further includes: a sealing head 5 and a sealing ring. The sealing heads 5 are arranged at both ends of the annular visual inspection robot, and sealing is achieved by adding sealing rings between the sealing heads 5 and the cylinder of the annular visual inspection robot. The mileage wheel unit 4 is arranged on the sealing head 5.

[0055] Optionally, the sealing head 5 is made of aluminum alloy.

[0056] Furthermore, the sealing head 5 at the tail end is not only responsible for sealing, but also used to install an aviation plug so as to connect to an external line to realize data transmission.

[0057] Furthermore, a through hole is designed at the front sealing head 5 to accommodate the front lens of the camera so as to photograph the inner wall of the pipe. The through hole part adopts a sapphire lens with a thickness of 6 mm and is sealed by a sealing ring to protect the camera lens and internal system from the influence of high pressure.

[0058] Furthermore, an aluminum alloy plate is fixed inside the sealing head 5 by a hinge, and the position of the aluminum alloy plate is adjusted to ensure that the camera is installed in the center. The two are respectively installed on both sides of the aluminum alloy plate to realize image acquisition and data processing, and the tail end is connected to a circular support by a hinge.

[0059] In the present invention, by providing sealing heads 5 and sealing rings at both ends of the annular visual inspection robot, the sealing of the device can be effectively ensured to prevent external liquids, gases or impurities from entering the interior of the device, thereby protecting the safe operation of the internal precision components in a high pressure or corrosive environment. At the same time, the mileage wheel unit 4 is integrated into the sealing head 5, which optimizes the structural layout of the device, enhances the overall durability and stability, and provides reliable protection for detection in complex environments.

[0060] In a possible implementation, the annular visual inspection robot further includes: an eddy current inspection unit. The eddy current inspection unit is connected to the battery unit 3. In the forward direction of the annular visual inspection robot, the eddy current inspection unit is located in front and the imaging unit 1 is located in the back. When the eddy current inspection unit detects that there is an abnormality on the wall surface of the oil and gas pipeline, a start signal of the imaging unit 1 is triggered, and the imaging unit 1 captures an image of the inner surface of the oil and gas pipeline.

[0061] In the present invention, by integrating an eddy current detection unit in the annular visual inspection robot and setting it at the front end of the forward direction, a rapid pre-screening of abnormalities on the pipeline wall can be achieved. When an abnormal area is detected, the eddy current unit can trigger the imaging unit 1 in time to take a focused shot, thereby improving the inspection efficiency, reducing unnecessary image acquisition and energy consumption, and providing more accurate high-resolution images for subsequent defect analysis, thereby improving the accuracy and reliability of the overall inspection.

[0062] In a possible implementation, the annular visual inspection robot further includes: a cleaning unit. The cleaning unit is connected to the battery unit 3. In the forward direction of the annular visual inspection robot, the cleaning unit is located in front and the imaging unit 1 is located in the rear. After the cleaning unit cleans the oil stains on the inner surface of the oil and gas pipeline, the imaging unit 1 captures the image of the inner surface of the oil and gas pipeline.

[0063] In the present invention, by integrating a cleaning unit in the annular visual inspection robot and arranging it at the front end in the forward direction, the inner wall of the pipeline can be cleaned before image acquisition to effectively remove oil stains and impurities. This not only ensures that the image captured by the imaging unit 1 is clearer, but also improves the accuracy of defect detection and avoids misjudgment due to surface contamination. At the same time, this design reduces the complexity of post-data processing and improves the overall efficiency and reliability of detection.

[0064] Refer to the instruction manual Figure 3 , showing a schematic structural diagram of a pressure difference rotary cleaning unit provided in an embodiment of the present invention.

[0065] In a possible implementation, the annular vision detection robot further includes: a leather cup driving unit 6 and a differential pressure rotary cleaning unit 7. The pressure difference between the rear end and the front end of the leather cup driving unit 6 drives the annular vision detection robot to move in the pipeline. The differential pressure rotary cleaning unit 7 includes: a base 71, a rotary fan blade 72, and a rotary glass cylinder 73. An air inlet hole 711 and an air outlet hole 712 are provided on the base 71, and an air flow channel is provided between the air inlet hole 711 and the air outlet hole 712. A rotary fan blade 72 is provided in the air flow channel. The rotary fan blade 72 is fixedly connected to the rotary glass cylinder 73. The air flow at the rear end of the leather cup driving unit 6 enters the air flow channel, drives the rotary fan blade 72 to rotate, and then drives the rotary glass cylinder 73 to rotate, throwing off the stains on the outer side of the rotary glass cylinder 73.

[0066] In the present invention, the air flow during the movement of the robot is ingeniously utilized, without the need for additional energy supply, to achieve the automatic cleaning function of the device. This not only keeps the glass cylinder clean and the vision clear, but also reduces energy consumption and structural complexity, improving the reliability and detection efficiency of the device operation.

[0067] Refer to the attached Figure 4 description, which shows a schematic flow chart of a method for detecting inner surface defects of an oil and gas pipeline provided by an embodiment of the present invention.

[0068] In a possible implementation, the processing unit 2 is specifically configured to:

[0069] S1: Obtain the annular image of the inner surface of the oil and gas pipeline captured by the imaging unit 1.

[0070] S2: Perform geometric correction on the annular image of the inner surface of the oil and gas pipeline.

[0071] Specifically, the pixel points of the annular image adopt polar coordinates while the actual image needs to be presented in a rectangular coordinate system. The farther the image point is from the center, the more serious the distortion is, and it needs to be corrected to the actual rectangular coordinate system. Therefore, the conversion from polar coordinates to rectangular coordinates and the distortion compensation technology can be used to correct the annular image. Convert the annular image from the polar coordinate system to the rectangular coordinate system to achieve flattening correction. By expanding the polar coordinates into rectangular coordinates, the annular image can be "strained" into a two-dimensional expanded view. The annular image may have radial distortion due to the reflection of the conical mirror, and it needs to be corrected through a distortion model. The distortion correction formula is:

[0072]

[0073] Where r c represents the corrected radial distance, r m represents the actually measured pixel point radial distance, k 1 and k 2 represent the radial distortion coefficients.

[0074] Furthermore, geometric correction can be achieved through a computer vision library (such as OpenCV).

[0075] S3: Based on the annular image of the inner surface of the oil and gas pipeline after geometric correction, defect detection of the inner surface of the oil and gas pipeline is carried out through a convolutional neural network.

[0076] Specifically, the present invention can adopt an improved YOLOv5 network structure to achieve defect detection of the inner surface of the oil and gas pipeline.

[0077] The traditional YOLOv5 network structure mainly includes: input, backbone network, neck network, and head network. The image of the inner surface of the oil and gas pipeline is input, the backbone network is used to extract the image features of the inner surface of the oil and gas pipeline, the neck network is used to fuse the extracted features, and the head network is used to perform defect detection based on the fused features.

[0078] The backbone network mainly adopts the CSPDarknet53 structure. Based on the Darknet53 of the YOLOv5 backbone network, the idea of CSPNet is borrowed, and a backbone structure with 5 CSP modules is designed. The convolution kernel size in front of each CSP module is 3×3, and the stride is 2 for downsampling. The CSP module first divides the feature map of the basic layer into two parts, and then merges them in a cross-stage manner, which can reduce the computational amount while ensuring the accuracy. Therefore, adopting the CSP network structure in the YOLOv5 backbone network has the advantages of enhancing the network learning ability, reducing the computational bottleneck and memory cost.

[0079] In order to further improve the feature extraction ability of the backbone network, the present invention adds a recursive gated convolution module to the backbone network. The recursive gated convolution module consists of a standard convolution, a linear mapping, and an element-wise multiplication. And layer normalization is added after the recursive gated convolution module. During the layer normalization process, the mean and variance of all channels are calculated and then normalized. The recursive gated convolution module adjusts the number of channels of the incoming feature map through two convolutional layers, and then divides the output features of the depthwise separable convolution into multiple parts. Each part performs an element-wise multiplication operation with the previous part, and finally the output features are obtained. Through the element-wise multiplication and recursive design, the interaction and fusion of high-order and low-order information of the feature map are realized, so that the information contained in the feature map is more abundant, the gradient diffusion phenomenon is reduced, and the feature extraction ability of the network is enhanced. The recursive operation is realized by continuously performing element-wise multiplication.

[0080] Furthermore, the recursive gated convolution module is specifically:

[0081]

[0082] Among them, x represents the input feature map, represents a linear projection mapping for mixing channel information, p 0 , q 0 represent two projected features segmented from the input feature map, and f represents a depth convolution operation. represents element-wise multiplication, p 1 represents the interaction feature. represents an inverse linear projection mapping for remapping the interaction feature back to the original channels of the input.

[0083] Furthermore, a recursive design can be introduced for multi-order mapping:

[0084]

[0085] Among them, p 0 , q 0 , …, q n-1 represent a set of projected features segmented from the input feature map.

[0086] Then, gated convolution is recursively performed:

[0087]

[0088] Among them, p k+1 represents the (k + 1)-th order interaction feature, f k represents the k-th order depth convolution operation, q k represents the k-th order projected feature, g k represents a linear mapping for matching the k-th order feature dimension, and α represents a scaling factor.

[0089]

[0090] Among them, Identity represents the identity function, Linear represents the linear mapping function, C k-1 represents the number of feature channels of the (k - 1)-th order, C k represents the number of feature channels of the k-th order, C represents the number of feature channels of the input feature map, n represents the total number of orders of the recursive gated convolution, and β represents the channel reduction factor, generally taken as 2.

[0091] In the present invention, by adding a recurrent gated convolution module to the backbone network and combining layer normalization and multi-order recurrence design, the feature extraction ability of the network can be effectively improved. The recurrent gated convolution module realizes the interaction and fusion of high-order and low-order information of the feature map through element-wise multiplication, making the feature expression more abundant, while reducing the gradient diffusion phenomenon and enhancing the stability of the model. Layer normalization further standardizes the feature distribution and optimizes the network training process. In addition, through multi-order recurrence design and gradual adjustment of the number of channels, deep interaction of features and efficient integration of information are achieved. While ensuring that the computational cost is controllable, the accuracy of feature extraction and the model's representation ability for complex patterns are greatly improved.

[0092] In order to further improve the feature extraction ability of the backbone network for the inner surface defects of oil and gas pipelines and at the same time suppress the interference of useless features, the present invention introduces a coordinate attention mechanism, which takes into account the relationship between position information and channels. It can not only capture cross-channel information, but also capture direction-aware and position-aware information, enabling the model to more accurately locate and identify the target area.

[0093] The coordinate attention mechanism decomposes global pooling into two one-dimensional feature encoding operations. Given the input X, pooling kernels of size (H,1) and (1,W) are used to encode each channel along the horizontal and vertical coordinates respectively. Then the coordinate information embedding transformation is specifically as follows:

[0094]

[0095] where z h represents the horizontally aware feature map after horizontal pooling, represents the feature map of the c-th channel after horizontal pooling, x c represents the input feature map in the c-th channel, h represents the channel height, i represents the width direction index value, W represents the feature map width, z w represents the feature map after vertical pooling, represents the vertically aware feature map of the c-th channel after vertical pooling, w represents the channel width, j represents the height direction index value, and H represents the feature map height.

[0096] Through the above two transformation operations, a pair of direction (X direction and Y direction) aware feature maps are obtained.

[0097] The X-direction and Y-direction aware feature maps are input into a shared 1×1 convolution transformation function to generate an intermediate feature map:

[0098]

[0099] where f represents the intermediate feature map, δ represents the non-linear activation function, F 1Denote the 1×1 convolution transformation function as z h Denote the horizontally perceived feature map as z w Denote the vertically perceived feature map.

[0100] Decompose the intermediate feature map along the spatial dimension into two independent tensors to obtain two independent feature maps.

[0101] Use two 1×1 convolutions to convert the two independent feature maps into the same number of channels as the input respectively:

[0102]

[0103] Among them, g h Denote the horizontal direction attention weight map as g w Denote the vertical direction attention weight map, σ denotes the Sigmoid activation function, and F h Denote the 1×1 convolution function in the horizontal direction, Denote the horizontally independent feature map obtained by decomposition as F w Denote the 1×1 convolution function in the vertical direction, Denote the vertically independent feature map obtained by decomposition.

[0104] Process the input feature map according to the attention weight map:

[0105]

[0106] Among them, y c Denote the output feature map after introducing coordinate attention for the c-th channel, Denote the horizontal direction attention weight map for the c-th channel, Denote the vertical direction attention weight map for the c-th channel.

[0107] In the present invention, by introducing the coordinate attention mechanism into the backbone network, the network's ability to extract the features of the inner surface defects of oil and gas pipelines is effectively improved, and at the same time, the interference of useless features is suppressed. The coordinate attention mechanism combines the position information and the relationship between channels, captures the direction-aware and position-aware information by decomposing the global pooling into one-dimensional feature encodings in the horizontal and vertical directions. Through the generation and fusion of the horizontal and vertical direction attention weights, the network can more accurately locate and identify the target area, enhance the accuracy and pertinence of the feature expression, and enable the model to have stronger robustness and performance in complex detection tasks.

[0108] The neck network adopts an FPN+PAN structure. Among them, the FPN (Feature Pyramid Network) layer transmits strong semantic features from top to bottom, and the PAN (Path Aggregation Network) layer transmits strong localization features from bottom to top, aggregating features for different detection layers at different backbone layers to improve the feature extraction ability.

[0109] The head network can use the Softmax activation function for defect classification.

[0110] Optionally, the main defect categories include: deformation, dislocation, breakage, corrosion, oxide layer spalling, sediment, and penetration, etc.

[0111] In order to improve the fault detection accuracy of the convolutional neural network, it is necessary to optimize the network parameters of the convolutional neural network. The traditional method uses the gradient descent method for optimization. However, the optimization result of the gradient descent method usually depends on the selection of the initial parameters. Poor initial values may lead to low optimization efficiency or unsatisfactory results. And the gradient descent method is prone to falling into local optimal solutions in high-dimensional non-convex optimization problems and cannot find the global optimal solution. Therefore, the present invention uses an improved heuristic algorithm (Honey Badger Algorithm) to optimize the network parameters of the convolutional neural network.

[0112] Among them, the Honey Badger Algorithm (HBA) is an intelligent optimization algorithm inspired by the foraging behavior of honey badgers. By simulating the olfactory tracking and digging strategies of honey badgers in the process of searching for prey, it conducts global search and local optimization in the solution space. This algorithm combines the balance mechanism of exploration and exploitation, has a strong ability to jump out of local optima, and is suitable for solving complex non-linear optimization problems.

[0113] Specifically, a loss function for detecting inner surface defects of oil and gas pipelines can be constructed for the convolutional neural network. The loss function can adopt an object box loss function, a cross-entropy loss function, a confidence loss function, etc.

[0114] Furthermore, the fitness function of the honey badger optimization algorithm is constructed with the loss function.

[0115] The honey badger individuals are initialized using Sine chaotic mapping. Each honey badger individual represents a set of feasible network parameters. Each honey badger individual consists of multiple dimensional components, and each component represents a network parameter:

[0116]

[0117] where x i represents the initial position of the i-th honey badger individual, lb i represents the lower bound of the i-th feasible solution, ubi represents the upper bound of the i-th feasible solution, y i represents the chaotic number corresponding to the i-th honey badger individual, y i-1 represents the chaotic number corresponding to the (i - 1)-th honey badger individual, μ represents the chaos parameter, generally taking 0.99.

[0118] In the present invention, by utilizing the ergodicity and initial sensitivity of the chaotic mapping, it helps to generate more uniformly distributed and wider coverage initial solutions in the solution space, thereby enhancing the diversity of the population and avoiding falling into local optimal solutions. At the same time, the Sine mapping is simple and efficient, and can improve the global search ability and convergence performance of the algorithm at a lower computational cost, laying a good foundation for the subsequent optimization process.

[0119] Adopt an elite selection strategy to retain the first half of the honey badger individuals with higher fitness values and discard the other half of the honey badger individuals to filter the population.

[0120] In the present invention, adopting the elite selection strategy to retain the first half of the honey badger individuals with higher fitness values can effectively improve the overall quality of the population and accelerate the convergence speed of the algorithm. By survival of the fittest, retaining the better-performing individuals helps to inherit excellent characteristics and guide the search direction, reducing the interference of low-quality individuals to the optimization process, thereby improving the search efficiency and stability of the global optimal solution.

[0121] In the mining stage, introduce a random number r 1 , and make a parallel selection between searching around the global optimal individual or the current individual to update the individual position:

[0122]

[0123] where represents the position of the i-th honey badger individual at the (t + 1)-th iteration, ω t represents the non-linear weight factor at the t-th iteration, represents the position of the i-th honey badger individual at the t-th iteration, x best represents the position where the global optimal individual is located, F represents the search direction control parameter, β represents the food acquisition ability of the honey badger individual, generally taking a fixed value of 6, I i represents the intensity factor of the i-th honey badger individual, α represents the density factor, r 1 , r 2 , r 3 and r 4 represent random numbers between 0 and 1.

[0124] In the present invention, introducing the random number r 1 in the mining stage and making a parallel search between the global optimal individual and the current individual helps to achieve a dynamic balance between local exploitation and global exploration. When When [condition], the individual mainly fine-tunes around itself to enhance the local search ability. When When [condition], the individual approaches the global optimal position to strengthen the global search ability. This mechanism can improve the convergence speed of the algorithm while effectively avoiding falling into local optimal solutions, enhancing the robustness and optimization efficiency of the algorithm. In addition, introducing non-linear perturbations using sine and cosine functions further increases the diversity of the search path and improves the adaptability to complex optimization problems.

[0125]

[0126] Among them, t represents the current iteration number, and T represents the maximum iteration number.

[0127] In the present invention, a non-linear weight factor is adopted, which can gradually reduce the weight during the iteration process, so that in the initial stage of the algorithm, the search process is more exploratory and can widely search the solution space. In the later stage, as the number of iterations increases, the weight gradually decreases, and the algorithm conducts more local development to finely optimize the quality of the solution. This weight decay strategy can balance the global search and local optimization capabilities, avoid premature convergence, and ensure finding the global optimal solution or a solution closer to the global optimum.

[0128]

[0129] Among them, r 5 represents a random number between 0 and 1, and S represents the concentration intensity.

[0130] In the present invention, the idea of inverse distance is introduced into the intensity factor, so that individuals closer to the global optimal solution have a higher "perception" intensity and thus participate in the search more actively. Combining the position relationship of the global optimal individual position, the current individual position, and its neighbors, the perception intensity of the individual to the target is dynamically measured.

[0131]

[0132] Among them, C represents the density constant, and exp represents the exponential function with the natural constant as the base.

[0133] In the present invention, the density factor can make the density gradually decay with the number of iterations, so as to maintain a larger perturbation range in the initial stage of the algorithm and enhance the global exploration ability. As the iteration progresses, the perturbation gradually decreases, prompting the algorithm to focus more on local fine search in the later stage. Such an exponential decay mechanism helps to implement the "exploration first, exploitation later" strategy in the optimization process, improve the overall convergence speed and optimization accuracy, and enhance the ability of the algorithm to jump out of local optima and approach the global optimal solution.

[0134]

[0135] where r 6 represents a random number between 0 and 1.

[0136] In the honey collection stage, update the individual position:

[0137]

[0138] where r 7 represents a random number between 0 and 1.

[0139] In the present invention, during the honey collection stage, the honey badger individuals can be appropriately adjusted according to the global optimal solution while maintaining exploration. The random factor r 7 provides more search diversity, preventing individuals from relying too much on the global optimal solution and resulting in premature convergence. Through the gradually decreasing density factor, the algorithm can gradually reduce the perturbation during the search process, enhance the local search ability, and finally improve the convergence accuracy and the ability to find the global optimal solution.

[0140] Update the fitness values of each honey badger individual and the global optimal individual.

[0141] Judge whether the current iteration number has reached the maximum iteration number. If so, output the set of network parameters represented by the honey badger individual with the highest current fitness. Otherwise, return to continue the iteration.

[0142] In the present invention, an improved honey badger optimization algorithm is used to optimize the network parameters of the convolutional neural network. Compared with the traditional gradient descent method, the honey badger optimization algorithm has stronger global search ability, can effectively avoid falling into local optimal solutions, and improve the reliability of the optimization results. It can significantly improve the optimization effect and application performance of the CNN, and provide more powerful model support for high-precision fault detection.

[0143] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A circular visual inspection robot for inner surface defect detection of oil and gas pipelines, characterized in that: include: An imaging unit (1), a processing unit (2), a battery unit (3) and a mileage wheel unit (4); The imaging unit (1), the processing unit (2), the battery unit (3) and the mileage wheel unit (4) are connected in series; The imaging unit (1) is used to capture an image of the inner surface of an oil and gas pipeline; The processing unit (2) is used to perform defect detection based on the captured image of the inner surface of the oil and gas pipeline; The battery unit (3) is used to supply power to the imaging unit (1) and the processing unit (2); The mileage wheel unit (4) is used to record the movement mileage; The imaging unit (1) comprises a camera (11), a conical mirror (12), and a glass tube (13); The glass cylinder (13) seals the imaging unit (1), and a containing space is formed inside the glass cylinder (13). The conical mirror (12) is arranged in the containing space. External light passes through the glass cylinder (13) to reach the conical mirror (12), and is projected into the camera (11) through the conical mirror (12), so that a ring-shaped image of the inner surface of the oil and gas pipeline is captured by the camera (11).

2. The annular visual inspection robot for inner surface defect detection of oil and gas pipelines according to claim 1 is characterized in that: The conical mirror (12) and the camera (11) are both centrally arranged.

3. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: The glass cylinder (13) is made of sapphire material.

4. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: The processing unit (2) comprises: a computer, a power control board and an LED control board; The power control board is connected to the battery unit (3); the power control board and the imaging unit (1) are both connected to the computer; the imaging unit (1), the mileage wheel unit (4), the LED control board and the LED light are all connected to the power control board; the LED control board is used for the LED light flashing function.

5. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: The mileage wheel unit (4) comprises: a mileage wheel (41), a mileage wheel base (42), a mileage wheel bracket (43), a rotating shaft, a magnet and a magnetic encoder; The mileage wheel bracket (43) is connected to the mileage wheel base (42), and the mileage wheel (41) is arranged on the mileage wheel bracket (43) via the rotating shaft; The magnet with radial magnetization is installed on the rotating shaft, and the magnet is located at the geometric center of the rotating shaft; When the annular visual inspection robot runs in a pipeline, the mileage wheel (41) contacts the inner wall of the pipeline, driving the rotating shaft connected by a key to rotate. When the rotating shaft rotates, the magnet rotates along with the rotating shaft, and the magnetic encoder senses the position change of the magnet and generates a pulse signal. The pulse signal triggers the camera (11) unit to capture an image.

6. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: Also includes: Sealing head (5) and sealing ring; The sealing heads (5) are arranged at both ends of the annular visual inspection robot, and sealing is achieved by adding the sealing ring between the sealing head (5) and the cylinder of the annular visual inspection robot; The mileage wheel unit (4) is arranged on the sealing head (5).

7. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: Also includes: an eddy current detection unit; The eddy current detection unit is connected to the battery unit (3); In the forward direction of the annular visual inspection robot, the eddy current inspection unit is located in front and the imaging unit (1) is located in the rear; When the eddy current detection unit detects that an abnormality exists on the wall surface of the oil and gas pipeline, a start signal of the imaging unit (1) is triggered, and an image of the inner surface of the oil and gas pipeline is captured by the imaging unit (1).

8. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1 is characterized in that: Also includes: Cleaning unit; The cleaning unit is connected to the battery unit (3); In the forward direction of the annular visual inspection robot, the cleaning unit is located in front and the imaging unit (1) is located in the rear; After the cleaning unit cleans the oil stains on the inner surface of the oil and gas pipeline, the imaging unit (1) captures an image of the inner surface of the oil and gas pipeline.

9. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1, characterized in that: Also includes: A leather cup driving unit (6) and a pressure differential rotation cleaning unit (7); The annular visual inspection robot is driven to move in the pipeline by a pressure difference between the rear end and the front end of the leather cup drive unit (6); The pressure difference rotary cleaning unit (7) comprises: a base (71), a rotary fan blade (72) and a rotary glass cylinder (73); An air inlet hole (711) and an air outlet hole (712) are provided on the base (71), and an air flow channel is provided between the air inlet hole (711) and the air outlet hole (712); The rotating fan blade (72) is arranged in the air flow channel; The rotating fan blade (72) is fixedly connected to the rotating glass cylinder (73); The airflow at the rear end of the leather cup driving unit (6) enters the airflow channel, driving the rotating fan blades (72) to rotate, thereby driving the rotating glass cylinder (73) to rotate, thereby removing the dirt on the outside of the rotating glass cylinder (73).

10. The annular visual inspection robot for inner surface defect inspection of oil and gas pipelines according to claim 1, characterized in that: The processing unit (2) is specifically used for: S1: Acquiring a ring-shaped image of the inner surface of the oil and gas pipeline photographed by the imaging unit (1); S2: performing geometric correction on the annular image of the inner surface of the oil and gas pipeline; S3: Based on the geometrically corrected annular image of the inner surface of the oil and gas pipeline, defects on the inner surface of the oil and gas pipeline are detected through a convolutional neural network.

Images